This project aims to understand the role of neighborhood residential trajectories for causing adolescent mental health and substance use problems, to inform potential prevention strategies early in life. Since most adulthood substance use and mental disorders originate early in life, it is imperative to focus on childhood exposures. Although evidence links adverse neighborhood contexts to impaired development, most studies are observational and cross sectional, limiting causal inference and policy translation. We propose a rigorous secondary data analysis using data from an experimental study of housing vouchers that may inform neighborhood context as a cause of health problems among low income minority populations: the Moving to Opportunity (MTO) study. MTO allows strong causal inferences, and though it was a promising policy to disrupt the toxic effects of neighborhood poverty, health researchers have had limited access to the data. It is necessary to unpack this expensive, ambitious housing experiment to understand the mechanisms by which non-health policies may promote, or indeed harm, health. We propose new analyses to answer significant questions for the design of rental assistance policy, and for the etiology of neighborhood health effects. We will leverage MTO's experimental design to understand how households select and remain in neighborhoods, how residential trajectories across time are influenced by health, and how these residential trajectories, in turn, influence health. To operationalize high opportunity neighborhoods, we apply a Geography of Opportunity framework - a policy-relevant analytic tool that emphasizes regional context and the continuing significance of social inequality for constraining residential choice, and for influencing residential preferences. We will model the dynamic nature of neighborhoods as well as the spatial, temporal moves of low-income families. To do so, we will construct residential trajectories by merging over 46,000 MTO geocoded address spells with a mosaic of retail, survey, and administrative neighborhood data to operationalize high opportunity neighborhoods across time. We will test for factors that influence household residential trajectories using growth curves and latent class analysis. We will then apply causal mediation methods, including novel semiparametric weight- based techniques, to test whether and when residential trajectories mediate MTO effects on adolescent health, to inform life course timing and mechanisms of neighborhood exposures for health effects. This project builds on a productive ongoing interdisciplinary collaboration, and leverages team expertise in mental health and substance use, neighborhood effects, adolescent development, life course and causal methods. Answers to our research questions may inform residential selection bias, which may, in turn, inform how to design more rigorous observational neighborhood studies. Further, since housing choice vouchers constitute a large portion of federal affordable housing dollars, our results can inform policy, e.g., to tailor housing mobility programs to address client health problems, and help low income families sustain moves to high opportunity areas.